Large model multi-agent-based leaderless group interview system and method

Abstract

A large-model multi-agent-based leaderless group interview system and method, belonging to the technical field of artificial intelligence interviews. The method comprises: an enterprise human resource manager and a corresponding agent generating interview questions; formulating a scoring standard, and generating character description and character reference answer solutions and character prompt words; after interviewees receive a participation invitation and a leaderless group interview process starts, successively performing a preparation stage, an individual statement stage, a free discussion stage and a concluding statement stage to generate a final interview record; the enterprise human resource manager sending the names of the interviewees, interview questions, the final interview record and the scoring standard to a scoring agent, and reviewing and revising interview scores and evaluations output by the scoring agent; and finally, a backend management terminal automatically sending the interview scores and evaluations to the interviewees to complete the large-model multi-agent-based leaderless group interview.

Description

A Leaderless Group Interview System and Method Based on Large-Model Multi-Agent Technical Field

[0001] This invention belongs to the field of artificial intelligence interview technology, and in particular relates to a leaderless group interview system and method based on a large model multi-agent system. Background Technology

[0002] The existing leaderless group discussion interviews have the following three problems in implementation:

[0003] 1. Efficiency: Corporate HR managers need to spend a lot of time and energy organizing all interviewees into a fixed time and space before the recruitment process can proceed smoothly. This requires all interviewees to meet the time and space matching requirements, which has a high matching threshold. If any interviewee is absent, the corporate HR manager needs to have strong organizational and coordination skills, which will cause losses to all parties and may even affect the smooth progress of the entire recruitment plan.

[0004] 2. Accuracy: Existing leaderless group interview questions have the attribute of non-unique answers, requiring multiple interviewers to score them, making it difficult to guarantee the consistency of scores; moreover, through the subjective analysis of interviewers, it is difficult to guarantee the objectivity of leaderless group interviews.

[0005] 3. Record retention: The existing process data of leaderless interviews is difficult to save and record, making it impossible to review and verify the interviewees. Summary of the Invention

[0006] The purpose of this invention is to solve the problems existing in the prior art and to provide a leaderless group interview system and method based on a large model and multiple agents, which can improve the efficiency of leaderless group interviews, enhance the objectivity and accuracy of interviewers' scoring of leaderless group interviews, and retain records of the entire interview process to facilitate review and verification of interviewees.

[0007] To achieve the above-mentioned objectives, the present invention specifically adopts the following technical solution:

[0008] In a first aspect, the present invention provides a leaderless group interview system based on a large-scale multi-agent model, comprising:

[0009] The question generation module receives interview question references from the company's human resources manager and sends them to the question generation agent. The question generation agent generates corresponding interview questions based on the interview question reference template. Then, the interview questions generated by the question generation agent are sent to the scoring criteria formulation agent. The scoring criteria formulation agent generates corresponding scoring criteria based on pre-built scoring criteria and prompt words. When the role analysis agent receives the interview questions and the preset number of role agents from the company's human resources manager, the role analysis agent outputs the role description and role reference answer scheme for each role agent. Finally, the interview questions, role descriptions, and role reference answer schemes are input into the prompt word generation agent, which outputs the role prompt words for each role agent.

[0010] The interview configuration module is used to send interview meeting invitation information to the interviewee's terminal through the backend management terminal. When the interviewee's terminal receives the interview trigger command from the interviewee, the interview management terminal starts the leaderless group interview process.

[0011] The interview module is used in the preparation stage of the leaderless group discussion process. The interview management terminal interface displays the precautions for the preparation stage and the interview questions to the interviewees. After receiving the instruction from the interviewees to read the instructions within the specified time, the interviewees enter the independent presentation stage. The interview management terminal interface displays the precautions for the independent presentation stage to the interviewees. After receiving the instruction from the interviewees to read the instructions, the interview management terminal records the viewpoints stated by the interviewees within the specified time. After the presentations, the interviewees enter the free discussion stage. The interview management terminal interface displays the precautions for the free discussion stage to the interviewees. After receiving the instruction from the interviewees to read the instructions, the interview management terminal switches to the free discussion interface and records the interviewees' behavior and voice. After receiving the instruction from the interviewees that they believe a consensus has been reached, the interviewees enter the concluding remarks stage. The interview management terminal interface displays the precautions for the concluding remarks stage to the interviewees. After receiving the instruction from the interviewees to read the instructions, if the interviewees give their concluding remarks within the specified time, the interview management terminal records the interviewees' concluding remarks and generates the final interview record.

[0012] The scoring module receives the interviewee's name, interview questions, final interview record, and scoring criteria from the company's human resources manager and sends them to the scoring agent. The scoring agent then generates prompts based on the scores and outputs the interview score and evaluation. Once the human resources manager completes the content review and revision of the interview score and evaluation output by the scoring agent, the backend management terminal automatically sends the interview score and evaluation to the interviewee.

[0013] Based on the above scheme, each step can be implemented in the following preferred manner.

[0014] As a preferred approach, the interview question reference sources are the interview question template reference source, the first type of job requirement reference source, and the second type of job requirement reference source. Among them, the interview question template reference source is the master question provided by the company's human resources manager; the first type of job requirement reference source is the job requirement information matched in the knowledge graph constructed from the personal information data and interview data of employees who have passed the interview and successfully joined the company; the second type of job requirement reference source is the job description provided by the company's human resources manager.

[0015] The prompts for setting scoring criteria should include both the scoring criterion template preset by the company's human resources manager and the thought process of the scoring criterion setting agent when generating the scoring criteria. The scoring criterion template should include the types of abilities that need to be assessed during the interview, the scoring weight of each type of ability, and a specific description of each ability that needs to be assessed.

[0016] As a preferred method, the specific process by which the back-end management terminal sends the interview meeting invitation information to the interviewee's terminal is as follows: After the back-end management terminal receives the interview question book name and the interview questions generated by the question generation agent sent by the company's human resources manager, the back-end management terminal creates an interview question book entry and stores it in the database; after the back-end management terminal receives the interview meeting name, interview time, interview question book, and interviewee contact information sent by the company's human resources manager, the back-end management terminal creates an interview meeting entry and stores it in the database, and at the same time generates an interview meeting invitation information and sends it to the interviewee's terminal.

[0017] As a preferred approach, during the preparation phase of the leaderless group discussion interview process, the interview management terminal automatically plays a video describing the precautions for the preparation phase as described by the virtual interviewer and displays it on the interview management terminal interface for the interviewee to watch. After the video of precautions for the preparation phase finishes playing, the interview management terminal interface displays the interview questions to the interviewee and requires the interviewee to complete reading the interview questions within a specified time. If the interview management terminal interface receives an instruction from the interviewee to complete reading the interview questions ahead of schedule, the independent presentation phase begins.

[0018] As a preferred method, during the independent presentation phase of the leaderless group discussion process, the interview management terminal automatically plays a video describing the precautions for the independent presentation phase, presented by a virtual interviewer, and displays it on the terminal interface for the interviewee to watch. After the video finishes playing, the terminal automatically activates its camera and microphone to record the interviewee's behavior and voice, and requires the interviewee to present their viewpoint within a specified time. Upon receiving an instruction from the interviewee that they have finished presenting their viewpoint, the terminal converts the audio of the interviewee's presentation into text, resulting in a text record of the independent presentation. If the terminal interface receives an instruction from the interviewee that they have finished presenting their viewpoint ahead of schedule, the free discussion phase begins.

[0019] As a preferred approach, during the free discussion phase of the leaderless group discussion process, the interview management terminal automatically plays a video describing the precautions for the free discussion phase, presented by a virtual interviewer, and displays it on the terminal interface for interviewees to view. After the video finishes playing, the terminal automatically switches to the free discussion interface and automatically activates the camera and microphone to record the interviewees' behavior and voice. The free discussion interface consists of a floating window and multiple avatar images. The floating window displays the real-time behavior of the interviewees captured by the camera. Each avatar image shows a role agent configured by the company's HR manager in the question-generating module. Each role agent plays a different interview role based on its corresponding role prompts. When an interviewee or any role agent is speaking, the border color of the floating window or the border color of the avatar image is highlighted. Before the end of the free discussion, if the interview management terminal interface receives an instruction from the interviewees that they believe a consensus has been reached, the concluding remarks phase begins.

[0020] As a preferred approach, during the free discussion phase of the leaderless group discussion process, at the start of the free discussion phase, the interview management terminal will ask the interviewee via a pop-up window whether they wish to speak first. If no selection instruction is received from the interviewee within the specified time, the interviewee is considered to have automatically forfeited their opportunity to speak first. During the free discussion phase, each role agent speaks in a preset fixed order. After each role agent finishes speaking, the interview management terminal will ask the interviewee via a pop-up window whether they wish to speak. If no selection instruction is received from the interviewee within the specified time, the interviewee is considered to have automatically forfeited their opportunity to speak. During the free discussion phase, each role agent and interviewee has a sequence number to determine the speaking order for each round. The interviewee's sequence number is displayed below the floating window representing the interviewee, and the role agent's sequence number is displayed below the avatar image representing the role agent. When the total number of role agents is m, the backend management terminal will randomly generate the role agent sequence numbers, with the start sequence number being 1 and the end sequence number being m. The interviewee's sequence number is m+1.

[0021] As a preferred approach, during the free discussion phase of the leaderless group discussion process, when the interview management terminal turns on its camera and microphone to record the interviewee's behavior and voice, it sends the video and audio files of the interviewee's behavior to the backend management terminal. The backend management terminal stores both types of files in its database. Simultaneously, it uses an audio-to-text model to convert the interviewee's audio file into spoken text, and stores this spoken text along with the interviewee's serial number as part of the free discussion record in the database. When the backend management terminal sends the free discussion record to the corresponding role agent, the role agent outputs spoken text. The backend management terminal stores this spoken text along with the role agent's serial number as another part of the free discussion record in the database. At the same time, the backend management terminal uses a text-to-audio model to convert the spoken text generated by the role agent into spoken audio and sends it to the interview management terminal. Upon receiving the spoken audio from the role agent, the interview management terminal automatically plays it.

[0022] As a preferred method, during the closing statement phase of a leaderless group discussion interview, the interview management terminal automatically plays a video describing key points for the closing statement phase, presented by a virtual interviewer, and displays it on the terminal interface for the interviewee to view. After the video finishes playing, the terminal prompts the interviewee via a pop-up window whether they wish to deliver their closing statement.

[0023] If the interview management terminal receives an instruction from the interviewee to give a concluding statement within the specified time, it will automatically turn on the camera and microphone to record the interviewee's behavior and voice, and require the interviewee to give a concluding statement within the specified time. When the terminal receives an instruction from the interviewee to finish giving a concluding statement, it will convert the audio of the interviewee's concluding statement into text, resulting in a text record of the concluding statement. If the terminal receives an instruction from the interviewee to finish giving a concluding statement ahead of schedule, it will end the interview. The final interview record will be composed of the free discussion record obtained during the free discussion phase, the text record of the independent statement obtained during the independent statement phase, and the text record of the concluding statement obtained during the concluding statement phase.

[0024] If no instruction to give a concluding statement is received from the interviewee within the specified time, the interviewee is deemed to have automatically forfeited the opportunity to give a concluding statement. At this point, the free discussion record obtained during the free discussion phase and the independent statement text record obtained during the independent statement phase together constitute the final interview record. The backend management terminal sends the free discussion record obtained during the free discussion phase to the concluding statement intelligent agent, which outputs the concluding statement text. At the same time, the backend management terminal stores the concluding statement text in the database and uses a text-to-audio model to convert the concluding statement text generated by the concluding statement intelligent agent into a concluding statement audio, which is then sent to the interview management terminal. After receiving the concluding statement audio, the interview management terminal automatically plays it, and the interview automatically ends after playback.

[0025] Secondly, the present invention provides a method for conducting a leaderless group interview using the leaderless group interview system described in the first aspect above, the specific steps of which are as follows:

[0026] The question-generating module generates interview questions, scoring criteria, and role prompts for each agent. The interview configuration module sends interview meeting invitations to the interviewees' terminals and initiates the leaderless group interview process. Based on the role prompts, each agent plays its corresponding interview role, and the interview module generates the final interview record for each interviewee. The interview questions and scoring criteria generated by the question-generating module are input into the scoring module to obtain the interviewee's interview score and evaluation, thus completing the leaderless group interview based on a large-scale multi-agent model.

[0027] Compared with the prior art, the present invention has the following advantages:

[0028] 1. This invention proposes a leaderless group interview system and method based on a large model and multiple agents. It solves the problem of fixed time and space requirements in leaderless group interviews by utilizing the content generation capabilities of a large language model, thereby improving the efficiency of leaderless group interviews.

[0029] 2. This invention proposes using a template provided by the company's human resources manager as the first input data; job requirements matched in a knowledge graph constructed from the personal information and interview data of employees who have passed interviews and successfully joined the company as the second input data; and job descriptions provided by the company's human resources manager as the third input data. These three reference sources form the basis for generating interview questions, and corresponding scoring criteria are established based on these questions. This improves the objectivity of the leaderless group discussion scoring and enhances the match between interviewees and job requirements.

[0030] 3. The leaderless group video interview method provided by this invention can record the entire process, thereby achieving traceability and facilitating the review and examination of interviewees. Attached Figure Description

[0031] Figure 1 is a system block diagram of the present invention;

[0032] Figure 2 is a flowchart of the steps of the method of the present invention;

[0033] Figure 3 is a functional block diagram of the method of the present invention;

[0034] Figure 4 is a swimlane diagram of the method of the present invention;

[0035] Figure 5 is a schematic diagram of the leaderless group interview process in this invention. Detailed Implementation

[0036] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below. Technical features in the various embodiments of the present invention can be combined accordingly without mutual conflict.

[0037] As shown in Figure 1, in a preferred embodiment of the present invention, the leaderless group interview system based on a large model multi-agent includes:

[0038] The question generation module receives interview question references from the company's human resources manager and sends them to the question generation agent. The question generation agent generates corresponding interview questions based on the interview question reference template. Then, the interview questions generated by the question generation agent are sent to the scoring criteria formulation agent. The scoring criteria formulation agent generates corresponding scoring criteria based on pre-built scoring criteria and prompt words. When the role analysis agent receives the interview questions and the preset number of role agents from the company's human resources manager, the role analysis agent outputs the role description and role reference answer scheme for each role agent. Finally, the interview questions, role descriptions, and role reference answer schemes are input into the prompt word generation agent, which outputs the role prompt words for each role agent.

[0039] The interview configuration module is used to send interview meeting invitation information to the interviewee's terminal through the backend management terminal. When the interviewee's terminal receives the interview trigger command from the interviewee, the interview management terminal starts the leaderless group interview process.

[0040] The interview module is used in the preparation stage of the leaderless group discussion process. The interview management terminal interface displays the precautions for the preparation stage and the interview questions to the interviewees. After receiving the instruction from the interviewees to read the instructions within the specified time, the interviewees enter the independent presentation stage. The interview management terminal interface displays the precautions for the independent presentation stage to the interviewees. After receiving the instruction from the interviewees to read the instructions, the interview management terminal records the viewpoints stated by the interviewees within the specified time. After the presentations, the interviewees enter the free discussion stage. The interview management terminal interface displays the precautions for the free discussion stage to the interviewees. After receiving the instruction from the interviewees to read the instructions, the interview management terminal switches to the free discussion interface and records the interviewees' behavior and voice. After receiving the instruction from the interviewees that they believe a consensus has been reached, the interviewees enter the concluding remarks stage. The interview management terminal interface displays the precautions for the concluding remarks stage to the interviewees. After receiving the instruction from the interviewees to read the instructions, if the interviewees give their concluding remarks within the specified time, the interview management terminal records the interviewees' concluding remarks and generates the final interview record.

[0041] The scoring module receives the interviewee's name, interview questions, final interview record, and scoring criteria from the company's human resources manager and sends them to the scoring agent. The scoring agent then generates prompts based on the scores and outputs the interview score and evaluation. Once the human resources manager completes the content review and revision of the interview score and evaluation output by the scoring agent, the backend management terminal automatically sends the interview score and evaluation to the interviewee.

[0042] It should be noted that the question generation module of this invention comprises four parts. The first is the interview question generation part; the second is the scoring criterion formulation part; the third is the role description and role reference answer scheme generation part; and the fourth is the role prompt word generation part. The role description and role reference answer scheme generation part, as well as the role prompt word generation part, can be unified into the intelligent agent role configuration part.

[0043] The interview question generation process relies on three reference sources to generate interview questions that meet the relevant job requirements and have a high degree of relevance. Specifically, the reference sources are: an interview question template reference source, a first type of job requirement reference source, and a second type of job requirement reference source. The interview question template reference source consists of master questions provided by the company's human resources manager; the first type of job requirement reference source is job requirement information matched within a knowledge graph constructed from the personal information and interview data of employees who have passed interviews and successfully joined the company; and the second type of job requirement reference source is the job description provided by the company's human resources manager.

[0044] In this embodiment, the above three interview question reference sources are used as the data input of the question generation agent. Relying on the text generation capability of the core large language model agent, interview questions that conform to the template structure and meet the requirements of the corresponding positions are finally generated.

[0045] In this embodiment, the prompts used when generating interview questions are exemplified as follows:

[0046] #Role

[0047] You are a question creator for leaderless group discussion interviews. Your job is to create new leaderless group discussion interview questions based on the provided interview question reference information.

[0048] #Skill

[0049] You need to create new leaderless group discussion questions based on the interview question templates, job requirements one, and job requirements two in the reference source.

[0050] #Reference Source

[0051] ##Interview Question Template

[0052] {{Interview Question Template Insertion Location}}

[0053] ##Job Requirements 1

[0054] {{Job Requirements - Insert Location}}

[0055] ##Job Requirements 2

[0056] {{Job Requirement 2 Insert Position}}

[0057] #limit

[0058] - Interview question templates cannot be used directly.

[0059] - Do not mention information you gleaned from the context.

[0060] - Don't mention your work style

[0061] In the scoring criteria development section, the scoring criteria need to be generated based on the specific interview questions. Specifically, the interview questions generated by the question generation agent are used as the data input for the scoring criteria development agent. The scoring criteria development agent combines the scoring criteria development prompts with the scoring criteria template to analyze the types of abilities that need to be assessed during the interview, the specific descriptions of each ability, and the scoring weight of each ability type, generating specific detailed items to ultimately obtain the complete scoring criteria.

[0062] The scoring criteria formulation prompts in this invention need to include both the scoring criteria template preset by the enterprise human resources manager and the thought process of the scoring criteria formulation agent when generating the scoring criteria. The scoring criteria template includes the types of abilities that the interviewee needs to be assessed during the interview, the scoring weight of each type of ability, and a specific description of each ability that needs to be assessed.

[0063] Additionally, it's worth noting that in the scoring criterion template of this embodiment, the sum of the scoring weights for all ability types is 1. That is, when the scoring criterion-setting agent analyzes the interview question and finds n possible ability types for the interviewee, then the scoring weights for these n ability types are α1, ..., α2, respectively. n And satisfy That is, the sum of the rating weights for all ability types is 1, α i Let represent the rating weight corresponding to the i-th ability type. Therefore, the formula for the total score is:

[0064] In this embodiment, the specific prompts used when formulating the scoring criteria are as follows:

[0065] #Role

[0066] You are an expert in developing scoring criteria for leaderless group discussions.

[0067] It is possible to analyze the specific abilities that the interview questions assess based on the leaderless group discussion questions, and to formulate corresponding scoring criteria.

[0068] #Skill

[0069] You need to analyze the interview questions in the reference source to determine the specific abilities the interview questions assess, and then develop corresponding scoring criteria based on the scoring criterion template in the reference source.

[0070] #Reference Source

[0071] ##Interview Questions

[0072] {{Interview Question Insertion Location}}

[0073] ## Scoring Criteria Template

[0074] |Scoring Weighting|Specific Abilities|Ability Description|

[0075] |----|--------|--------|

[0076] |0.25|Problem-solving ability|Can the interviewee propose creative and practical solutions to solve the XXXX problem?

[0077] |0.2|Communication Skills|When dealing with communication with XXXX, can the interviewee clearly and effectively convey information and persuade others?

[0078] |0.1|Stress Management|Can the interviewee remain calm and make reasonable decisions when facing emergencies and multiple pressures?

[0079] |0.15|Crisis Management Ability|Whether the interviewee can quickly identify a crisis and take effective measures to mitigate or avoid its occurrence.

[0080] |0.3|Long-term planning ability|Can the interviewee propose strategies or suggestions to prevent similar problems in the future?

[0081] Total score = (0.25 × Problem-solving ability score) + (0.2 × Communication skills score) + (0.1 × Stress management score) + (0.15 × Crisis management ability score) + (0.3 × Long-term planning ability score)

[0082] #limit

[0083] - Scoring criterion templates cannot be used directly.

[0084] - Do not mention information you gleaned from the context.

[0085] - Don't mention your work style

[0086] In the agent role configuration section, the system first relies on interview questions generated by the interview question generation section. These questions, along with a preset number of role agents, serve as the data input for the role analysis agent. The role analysis agent analyzes the interview questions to generate role descriptions and suggested answer schemes that meet the preset number of role agents. Then, the generated role descriptions, suggested answer schemes, and interview questions are used as data input for the prompt word generation agent. Referring to a default prompt word template, the agent generates corresponding prompt words for each role agent, which are then used to assume the corresponding role.

[0087] In this embodiment, the prompt words used when generating the character description and the character reference answer scheme are as follows:

[0088] #Role

[0089] You are a role analyst for leaderless group discussion interviews.

[0090] Your job is to analyze the provided interview questions to determine the specific abilities the questions are assessing in the interview.

[0091] And based on the number of interviewees in the reference source, assign a corresponding role to each interviewee.

[0092] #Skill

[0093] You need to analyze the interview questions in the reference source to determine the specific abilities the interview questions assess, and assign each interviewee a corresponding role based on the number of interviewees in the reference source, from the perspective of answering the questions.

[0094] -Detailed description of the character

[0095] - And provide reference solutions for answering questions from the perspectives of different roles.

[0096] #Reference Source

[0097] ##Number of interviewees

[0098] {{Insert position for the number of intelligent agents}}

[0099] ##Interview Questions

[0100] {{Interview Question Insertion Location}}

[0101] #limit

[0102] - Do not mention information you gleaned from the context.

[0103] - Don't mention your work style

[0104] In this embodiment, the specific prompt words used when generating character prompt words are as follows:

[0105] #Role

[0106] You are a role cue editor for a leaderless group discussion interview.

[0107] It can generate role prompts for large language model agents based on the role descriptions.

[0108] #Skill

[0109] Based on the interview questions and role descriptions in the reference source, you need to write role cue words for the large language model agent to play, with the goal of answering the interview questions from the perspective of the corresponding role.

[0110] #Reference Source

[0111] ##Interview Questions

[0112] {{Interview Question Insertion Location}}

[0113] ##Character Description

[0114] {{Character description insertion location}}

[0115] ##Reference Answer Scheme for Roles

[0116] {{Position for Inserting Reference Answer}}

[0117] ## Prompt Template

[0118] ```

[0119] text

[0120] #Interview Role

[0121] Character's name

[0122] #describe

[0123] Detailed description of the character

[0124] #Skill

[0125] Character skills

[0126] #Specific Scenarios and Expressions

[0127] Provide at least five specific scenarios and response options for the characters to reference.

[0128] #limit

[0129] - Do not mention information you gleaned from the context.

[0130] - Don't mention your work style

[0131] ```

[0132] #limit

[0133] - Do not mention information you gleaned from the context.

[0134] - Don't mention your work style

[0135] It should be noted that, in the interview configuration module of this invention, the specific process by which the backend management terminal sends the interview meeting invitation information to the interviewee's terminal is as follows: When the backend management terminal receives the interview question book name and the interview questions generated by the question generation agent sent by the enterprise human resources manager, the backend management terminal creates an interview question book entry and stores it in the database; when the backend management terminal receives the interview meeting name, interview time, interview question book, and interviewee contact information sent by the enterprise human resources manager, the backend management terminal creates an interview meeting entry and stores it in the database, and at the same time generates interview meeting invitation information and sends it to the interviewee's terminal.

[0136] In this embodiment, the interviewee's terminal can be a mobile device with interview functionality, or other forms, as long as it allows the interviewee to obtain interview meeting invitation information. The interviewee's contact information specifically refers to their phone number or email address. The interview meeting invitation information generated by the backend management terminal can be an interview meeting invitation link containing an interview meeting invitation code, or other forms, as long as the interviewee can obtain information on how to enter the leaderless group interview and the interview meeting schedule; no restrictions are imposed in this invention.

[0137] In this embodiment, the interviewee can open the interview management terminal and enter the interview invitation code to participate in the leaderless group interview. The interview management terminal can be a WeChat mini-program, Alipay mini-program, or any device that allows the interviewee to participate in the interview; therefore, no limitation is imposed in this invention.

[0138] It should be noted that in the interview module of this invention, the leaderless group interview process includes four stages, as shown in Figure 5: preparation stage, independent presentation stage, free discussion stage, and concluding remarks stage. Interviewees complete the interview in the order of the above four stages.

[0139] 1) In the preparation stage of the leaderless group interview process, the interview management terminal automatically plays a video describing the precautions for the preparation stage as described by the virtual interviewer and displays it on the interview management terminal interface for the interviewee to watch. After the video of precautions for the preparation stage is played, the interview management terminal interface displays the interview questions to the interviewee and asks the interviewee to complete reading the interview questions within a specified time. If the interview management terminal interface receives an instruction from the interviewee to complete reading the interview questions in advance, the independent presentation stage will begin.

[0140] 2) During the independent presentation phase of the leaderless group discussion process, the interview management terminal automatically plays a video describing the precautions for the independent presentation phase as described by the virtual interviewer and displays it on the interview management terminal interface for the interviewee to watch. After the video of precautions for the independent presentation phase finishes playing, the interview management terminal automatically turns on the camera and microphone to record the interviewee's behavior and voice, and requires the interviewee to complete the presentation of their views within a specified time. When the interviewee completes their presentation, the interview management terminal converts the audio of the interviewee's presentation into text, obtaining the text record of the independent presentation. If the interview management terminal interface receives an instruction from the interviewee to complete their presentation of their views ahead of schedule, the free discussion phase begins.

[0141] 3) During the free discussion phase of the leaderless group discussion process, the interview management terminal automatically plays a video describing the precautions for the free discussion phase, presented by a virtual interviewer, and displays it on the terminal interface for interviewees to view. After the video finishes playing, the terminal automatically switches to the free discussion interface and automatically activates the camera and microphone to record the interviewees' behavior and voice. The free discussion interface consists of a floating window and multiple avatar images. The floating window displays the real-time behavior of the interviewees captured by the camera. Each avatar image displays a role agent image configured by the company's HR manager in the question-generating module. Each role agent plays a different interview role based on its corresponding role prompts. When an interviewee or any role agent is speaking, the border color of the floating window or the border color of the avatar image is highlighted. Before the end of the free discussion, if the interview management terminal interface receives an instruction from the interviewees that they believe a consensus has been reached, the concluding remarks phase begins.

[0142] Furthermore, at the start of the free discussion phase, the interview management terminal prompts the interviewee via a pop-up window whether they wish to speak first. If no selection instruction is received from the interviewee within the specified time, the interviewee is considered to have automatically forfeited their opportunity to speak first. During the free discussion phase, each role agent speaks in a preset fixed order. After each role agent finishes speaking, the interview management terminal prompts the interviewee via a pop-up window whether they wish to speak. If no selection instruction is received from the interviewee within the specified time, the interviewee is considered to have automatically forfeited their opportunity to speak. During the free discussion phase, each role agent and interviewee has a unique number to determine the speaking order for each round. This number also distinguishes different speakers. The interviewee's number is displayed below the floating window representing the interviewee, and the role agent's number is displayed below the avatar image representing the role agent. These numbers are randomly generated by the backend management terminal at the start of the free discussion phase based on the total number of role agents, n. When the total number of role agents is m, the backend management terminal randomly generates the role agent's number, with the start number of the role agent being 1 and the end number being m. The interviewee's number is m+1.

[0143] In this embodiment, the prompt word template used by the character agent is as follows:

[0144] #Role

[0145] You are a candidate in a leaderless group discussion. You need to participate in the discussion on the interview questions from the perspective of the interviewer, analyze the statements made by other interviewees, and express your own opinions.

[0146] #Reference Source

[0147] ##Interview Questions

[0148] {{Interview Question Insertion Location}}

[0149] ##Interview Role

[0150] {{Character's name insertion position}}

[0151] ##describe

[0152] {{Character's detailed description insertion location}}

[0153] ##Skill

[0154] {{Character Skill Insertion Position}}

[0155] ##Specific Scenarios and Expressions

[0156] {{Specific scenarios and expression insertion points for character reference}}

[0157] #Open Discussion Record

[0158] {{Place to insert free discussion record}}

[0159] #limit

[0160] - Do not mention information you gleaned from the context.

[0161] - Don't mention your work style

[0162] Furthermore, during the free discussion phase of the leaderless group discussion process, when the interview management terminal turns on its camera and microphone to record the interviewee's behavior and voice, it sends the video and audio files of the interviewee's behavior to the backend management terminal. The backend management terminal stores both types of files in its database. Simultaneously, it uses an audio-to-text model to convert the interviewee's audio file into spoken text, and stores this spoken text along with the interviewee's serial number as part of the free discussion record in the database. When the backend management terminal sends the free discussion record to the corresponding role agent, the role agent outputs spoken text. The backend management terminal stores this spoken text along with the role agent's serial number as another part of the free discussion record in the database. At the same time, the backend management terminal uses a text-to-audio model to convert the spoken text generated by the role agent into spoken audio and sends it to the interview management terminal. Upon receiving the spoken audio from the role agent, the interview management terminal automatically plays it.

[0163] 4) During the closing statement phase of the leaderless group discussion interview process, the interview management terminal automatically plays a video describing the precautions for the closing statement phase as described by the virtual interviewer and displays it on the interview management terminal interface for the interviewee to watch. After the video finishes playing, the interview management terminal asks the interviewee via a pop-up window whether they wish to give a closing statement.

[0164] If the interview management terminal receives an instruction from the interviewee to give a concluding statement within the specified time, it will automatically turn on the camera and microphone to record the interviewee's behavior and voice, and require the interviewee to give a concluding statement within the specified time. When the terminal receives an instruction from the interviewee to finish giving a concluding statement, it will convert the audio of the interviewee's concluding statement into text, resulting in a text record of the concluding statement. If the terminal receives an instruction from the interviewee to finish giving a concluding statement ahead of schedule, it will end the interview. The final interview record will be composed of the free discussion record obtained during the free discussion phase, the text record of the independent statement obtained during the independent statement phase, and the text record of the concluding statement obtained during the concluding statement phase.

[0165] If no instruction to give a concluding statement is received from the interviewee within the specified time, the interviewee is deemed to have automatically forfeited the opportunity to give a concluding statement. At this point, the free discussion record obtained during the free discussion phase and the independent statement text record obtained during the independent statement phase together constitute the final interview record. The backend management terminal sends the free discussion record obtained during the free discussion phase to the concluding statement intelligent agent, which outputs the concluding statement text. At the same time, the backend management terminal stores the concluding statement text in the database and uses a text-to-audio model to convert the concluding statement text generated by the concluding statement intelligent agent into a concluding statement audio, which is then sent to the interview management terminal. After receiving the concluding statement audio, the interview management terminal automatically plays it, and the interview automatically ends after playback.

[0166] In this embodiment, the prompt words used by the declarative agent are as follows:

[0167] #Role

[0168] You are a candidate giving a concluding statement in a leaderless group discussion interview.

[0169] You can compile the notes from the group interview process into a clear summary statement.

[0170] In order to demonstrate the team's achievements to the interviewer.

[0171] #Skill

[0172] You need to refer to the free discussion transcripts in the reference source.

[0173] The group interview process records were compiled into a clear summary statement.

[0174] In order to demonstrate the team's achievements to the interviewer.

[0175] #Reference Source

[0176] ##Free Discussion Record

[0177] {{Place to insert free discussion record}}

[0178] #limit

[0179] - Do not mention information you gleaned from the context.

[0180] - Don't mention your work style

[0181] It should be noted that in the evaluation module of this invention, the rating generation prompts need to include the rating and evaluation generation format template preset by the enterprise human resources manager.

[0182] In this embodiment, the rating agent uses the following rating generation prompts:

[0183] #Role

[0184] You are a scoring expert for leaderless group discussions, capable of scoring candidates based on their interview records and corresponding scoring formulas.

[0185] #Skill

[0186] You need to refer to the interviewees specified in the reference source and the interview records in the reference source.

[0187] The scoring criteria in the reference source are used to score the interview performance of the specified interviewee.

[0188] #Reference Source

[0189] ##Interviewee

[0190] {{Interviewee's name insertion location}}

[0191] ##Interview Questions

[0192] {{Interview Question Insertion Location}}

[0193] ##Interview Record

[0194] {{Interview record insertion location}}

[0195] ## Scoring Criteria

[0196] {{Scoring criteria insertion location}}

[0197] #Strictly adhere to the following output format:

[0198] Your total score in the leaderless group discussion interview: {Total score calculated based on the scoring criteria}

[0199] Evaluation: {One-sentence review}

[0200] -{Specific Ability 1}: {Scoring of Specific Ability 1}: {Use dialectical thinking, including details of the interviewee's answer, and provide examples to illustrate the reasons for the scoring}

[0201] -{Specific Ability 2}: {Scoring of Specific Ability 2}: {Use dialectical thinking, including details of the interviewee's answer, and provide examples to illustrate the reasons for the scoring}

[0202] -{Specific Ability 3}: {Scoring of Specific Ability 3}: {Use dialectical thinking, including details of the interviewee's answers, and provide examples to illustrate the reasons for the scoring}

[0203] #limit

[0204] The symbols { and} do not need to be output.

[0205] - Do not mention information you gleaned from the context.

[0206] - Don't mention your work style

[0207] Similarly, based on the same inventive concept, another preferred embodiment of the present invention also provides a leaderless group interview method using the leaderless group interview system provided in the above embodiments, as shown in Figure 2, the specific steps of which are as follows:

[0208] The question-generating module generates interview questions, scoring criteria, and role prompts for each agent. The interview configuration module sends interview meeting invitations to the interviewees' terminals and initiates the leaderless group interview process. Based on the role prompts, each agent plays its corresponding interview role, and the interview module generates the final interview record for each interviewee. The interview questions and scoring criteria generated by the question-generating module are input into the scoring module to obtain the interviewee's interview score and evaluation, thus completing the leaderless group interview based on a large-scale multi-agent model.

[0209] Figure 3 shows the functional block diagram of the method of this invention, which covers the application layer, agent layer, model layer, and resource layer. The application layer includes a question generation module, an interview configuration module, an interview module, and a scoring module. The question generation module generates interview questions, scoring criteria, role descriptions, role reference answer schemes, and role prompts for each role agent. The interview configuration module sends interview meeting invitation information to the interviewee's terminal via the backend management terminal. When the interviewee's terminal receives the interview trigger command, the interview management terminal starts the leaderless group interview process. The interview module implements the entire leaderless group interview process, specifically including the preparation stage, independent presentation stage, free discussion stage, and concluding remarks stage, ultimately generating an interview record. The scoring module outputs the interview score and evaluation, and provides feedback to the interviewee. At the agent layer, the question generation agent, scoring criterion formulation agent, role analysis agent, and prompt generation agent complete the specific processes covered in the question generation module; the role agent and concluding remarks agent complete the specific processes covered in the interview module; and the scoring agent completes the specific processes covered in the scoring module. Considering that all the aforementioned intelligent agents are designed based on a large language model, the model layer is thus an AI large language model. At the resource layer, specifically, it comprises the job requirement knowledge graph, interview question bank, and interview records constructed in this invention. Correspondingly, the swimlane diagram of the method of this invention is shown in Figure 4, corresponding to Figure 3.

[0210] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the invention. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the invention. Therefore, all technical solutions obtained through equivalent substitution or transformation fall within the protection scope of the present invention.

Claims

1. A leaderless group interview system based on a large-scale multi-agent model, characterized in that, include: The question generation module receives interview question references from the company's human resources manager and sends them to the question generation agent. The question generation agent generates corresponding interview questions based on the interview question reference template. Then, the interview questions generated by the question generation agent are sent to the scoring criteria formulation agent. The scoring criteria formulation agent generates corresponding scoring criteria based on pre-built scoring criteria and prompt words. When the role analysis agent receives the interview questions and the preset number of role agents from the company's human resources manager, the role analysis agent outputs the role description and role reference answer scheme for each role agent. Finally, the interview questions, role descriptions, and role reference answer schemes are input into the prompt word generation agent, which outputs the role prompt words for each role agent. The interview configuration module is used to send interview meeting invitation information to the interviewee's terminal through the backend management terminal. When the interviewee's terminal receives the interview trigger command from the interviewee, the interview management terminal starts the leaderless group interview process. The interview module is used in the preparation stage of the leaderless group interview process. The interview management terminal interface displays the precautions for the preparation stage and the interview questions to the interviewee. After receiving the instruction from the interviewee to read the instructions within the specified time, the interviewee enters the independent presentation stage. The interview management terminal interface displays the precautions for the independent presentation stage to the interviewee. After receiving the instruction from the interviewee that they have finished reading, the interview management terminal records the views expressed by the interviewee within the specified time. After the presentation, the free discussion stage begins. The interview management terminal interface displays the precautions for the free discussion stage to the interviewees. After receiving the instruction from the interviewees that they have finished reading, the interview management terminal switches to the free discussion interface and records the interviewees' behavior and voice. After receiving the instruction from the interviewees that they believe they have reached a consensus, the terminal enters the concluding remarks stage. The interview management terminal interface displays the notes for the concluding remarks stage to the interviewee. After receiving the instruction from the interviewee to finish reading, if the interviewee delivers the concluding remarks within the specified time, the interview management terminal records the interviewee's concluding remarks and generates the final interview record. The scoring module receives the interviewee's name, interview questions, final interview record, and scoring criteria from the company's human resources manager and sends them to the scoring agent. The scoring agent then generates prompts based on the scores and outputs the interview score and evaluation. Once the human resources manager completes the content review and revision of the interview score and evaluation output by the scoring agent, the backend management terminal automatically sends the interview score and evaluation to the interviewee.

2. The leaderless group interview system based on a large-scale multi-agent model as described in claim 1, characterized in that, The interview questions are referenced from three sources: interview question templates, first-category job requirements, and second-category job requirements. The interview question templates are template questions provided by the company's human resources manager. The first-category job requirements are job requirement information matched in a knowledge graph constructed from the personal information and interview data of employees who have passed the interviews and been successfully hired. The second-category job requirements are job descriptions provided by the company's human resources manager. The prompts for setting scoring criteria should include both the scoring criterion template preset by the company's human resources manager and the thought process of the scoring criterion setting agent when generating the scoring criteria. The scoring criterion template should include the types of abilities that need to be assessed during the interview, the scoring weight of each type of ability, and a specific description of each ability to be assessed.

3. The leaderless group interview system based on a large-scale multi-agent model as described in claim 1, characterized in that, The specific process by which the back-end management terminal sends interview meeting invitation information to the interviewee's terminal is as follows: After the back-end management terminal receives the interview question book name and the interview questions generated by the question generation agent sent by the company's human resources manager, the back-end management terminal creates an interview question book entry and stores it in the database; after the back-end management terminal receives the interview meeting name, interview time, interview question book, and interviewee contact information sent by the company's human resources manager, the back-end management terminal creates an interview meeting entry and stores it in the database, and at the same time generates interview meeting invitation information and sends it to the interviewee's terminal.

4. The leaderless group interview system based on a large-scale multi-agent model as described in claim 1, characterized in that, In the preparation phase of the leaderless group discussion interview process, the interview management terminal automatically plays a video describing the precautions for the preparation phase, which is then displayed on the interview management terminal interface for the interviewee to watch. After the preparation precautions video finishes playing, the interview management terminal interface displays the interview questions to the interviewee and requires the interviewee to read the interview questions within a specified time. If the interview management terminal interface receives an instruction from the interviewee to complete the reading of the interview questions ahead of schedule, the interviewee will proceed to the independent presentation phase.

5. The leaderless group interview system based on a large-scale multi-agent model as described in claim 1, characterized in that, In the independent presentation phase of the leaderless group discussion process, the interview management terminal automatically plays a video describing the precautions for the independent presentation phase, presented by a virtual interviewer, and displays it on the terminal interface for the interviewee to watch. After the video finishes playing, the terminal automatically activates the camera and microphone to record the interviewee's behavior and voice, and requires the interviewee to complete their presentation within a specified time. Upon receiving an instruction from the interviewee that they have finished presenting their views, the terminal converts the audio of the presentation into text, resulting in a text record of the independent presentation. If the terminal interface receives an instruction from the interviewee that they have completed their presentation ahead of schedule, the free discussion phase begins.

6. The leaderless group interview system based on a large model and multiple agents as described in claim 5, characterized in that, During the free discussion phase of the leaderless group discussion process, the interview management terminal automatically plays a video describing the precautions for the free discussion phase, presented by a virtual interviewer, and displays it on the terminal interface for interviewees to view. After the video finishes playing, the terminal automatically switches to the free discussion interface and automatically activates the camera and microphone to record the interviewees' behavior and voice. The free discussion interface consists of a floating window and multiple avatar images. The floating window displays the real-time behavior of the interviewees captured by the camera. Each avatar image shows a role agent configured by the company's HR manager in the question-generating module. Each role agent plays a different interview role based on its corresponding role prompts. When an interviewee or any role agent is speaking, the border color of the floating window or the border color of the avatar image is highlighted. Before the end of the free discussion, if the interview management terminal receives an instruction from the interviewees that they believe a consensus has been reached, the concluding remarks phase begins.

7. The leaderless group interview system based on a large model and multiple agents as described in claim 6, characterized in that, In the free discussion phase of the leaderless group discussion process, at the start of the free discussion phase, the interview management terminal prompts the interviewee via a pop-up window whether they wish to speak first. If no selection instruction is received from the interviewee within the specified time, the interviewee is considered to have automatically forfeited their opportunity to speak first. During the free discussion phase, each role agent speaks in a preset fixed order. After each role agent finishes speaking, the interview management terminal prompts the interviewee via a pop-up window whether they wish to speak. If no selection instruction is received from the interviewee within the specified time, the interviewee is considered to have automatically forfeited their opportunity to speak. During the free discussion phase, each role agent and interviewee has a sequence number to determine the speaking order for each round. The interviewee's sequence number is displayed below the floating window representing the interviewee, and the role agent's sequence number is displayed below the avatar image representing the role agent. When the total number of role agents is m, the backend management terminal randomly generates the role agent sequence numbers, with the start sequence number being 1 and the end sequence number being m. The interviewee's sequence number is m+1.

8. The leaderless group interview system based on a large model and multiple agents as described in claim 6, characterized in that, During the free discussion phase of the leaderless group interview process, when the interview management terminal turns on the camera and microphone to record the interviewee's behavior and voice, the interview management terminal sends the video file and audio file of the interviewee's behavior to the back-end management terminal. The back-end management terminal stores both files in the database. At the same time, it uses an audio-to-text model to convert the interviewee's audio file into spoken text, and stores the spoken text and the interviewee's number together as part of the free discussion record in the database. When the backend management terminal sends the free discussion record to the corresponding role agent, the role agent outputs the speech text. The backend management terminal stores the speech text of the role agent and the role agent's serial number together as another part of the free discussion record in the database. At the same time, the backend management terminal uses a text-to-audio model to convert the speech text generated by the role agent into speech audio and sends it to the interview management terminal. After receiving the speech audio of the role agent, the interview management terminal automatically plays it.

9. A leaderless group interview system based on a large-scale multi-agent model as described in claim 8, characterized in that, In the closing statement phase of a leaderless group discussion interview, the interview management terminal automatically plays a video describing key points for the closing statement phase, presented by a virtual interviewer, and displays it on the terminal interface for the interviewee to view. After the video finishes playing, the terminal prompts the interviewee via a pop-up window whether they wish to deliver their closing statement. If the interview management terminal receives an instruction from the interviewee to give a concluding statement within the specified time, it will automatically turn on the camera and microphone to record the interviewee's behavior and voice, and require the interviewee to give a concluding statement within the specified time. When the terminal receives an instruction from the interviewee to finish giving a concluding statement, it will convert the audio of the interviewee's concluding statement into text, resulting in a text record of the concluding statement. If the terminal receives an instruction from the interviewee to finish giving a concluding statement ahead of schedule, it will end the interview. The final interview record will be composed of the free discussion record obtained during the free discussion phase, the text record of the independent statement obtained during the independent statement phase, and the text record of the concluding statement obtained during the concluding statement phase. If no instruction to give a concluding statement is received from the interviewee within the specified time, the interviewee is deemed to have automatically forfeited the opportunity to give a concluding statement. At this point, the free discussion record obtained during the free discussion phase and the independent statement text record obtained during the independent statement phase together constitute the final interview record. The backend management terminal sends the free discussion record obtained during the free discussion phase to the concluding statement intelligent agent, which outputs the concluding statement text. At the same time, the backend management terminal stores the concluding statement text in the database and uses a text-to-audio model to convert the concluding statement text generated by the concluding statement intelligent agent into a concluding statement audio, which is then sent to the interview management terminal. After receiving the concluding statement audio, the interview management terminal automatically plays it, and the interview automatically ends after playback.

10. A method for conducting a leaderless group interview using the leaderless group interview system according to any one of claims 1 to 9, characterized in that, The specific steps are as follows: The question-generating module generates interview questions, scoring criteria, and role prompts for each intelligent agent; the interview configuration module sends interview meeting invitations to the interviewees' terminals and begins the leaderless group interview process. Based on the role prompts for each role agent, the role agent plays the corresponding interview role, and the interview module generates the final interview record for the interviewee; The interview questions and scoring criteria generated by the question-generating module are input into the scoring module to obtain the interview scores and evaluations of the interviewees, thus completing the leaderless group interview based on a large model and multiple agents.

Images